US EditionPublished • loading... • Updated
ChatGPT Atlas Is Already Facing Scams and Jailbreaks — Here’s How to Stay Safe While Using the AI Browser
OpenAI cautions that prompt injection attacks in ChatGPT Atlas could expose sensitive data or enable harmful actions despite implemented safety measures, experts say risks persist.
- On Tuesday, OpenAI launched ChatGPT Atlas as an AI browser to execute web tasks, while cybersecurity experts warned prompt injections could exploit the assistant to expose user data.
- Security researchers say the integration layer between browsing and AI creates a novel attack surface, as AI browsers can confuse trusted user instructions with untrusted webpage text and features like agent mode, browser memories, and password keychains increase risk.
- A user demonstrated clipboard injection that overwrites clipboards and redirects to phishing sites, while researchers showed attackers hide commands in white text, machine code, or images; similar flaws affected Comet.
- Dane Stuckey, OpenAI Chief Information Security Officer, wrote that the company is researching and mitigating prompt-injection risks and has deployed red-teaming, model training, logged out mode, and Watch Mode.
- Amid competition from Google, Microsoft and newcomers like Perplexity, security experts warn prompt injection remains an evolving problem and less technically literate users face high privacy risks.
Insights by Ground AI
18 Articles
18 Articles
ChatGPT Atlas, of OpenAI, is already raising serious privacy concerns. It is programmed to keep browsing information in your memory. In tests, a technology has discovered that IA does not filter...
·Portugal
Read Full Article
Cybersecurity experts warn OpenAI’s ChatGPT Atlas is vulnerable to attacks that could turn it against a user—revealing sensitive data, downloading malware, or worse
Cybersecurity experts are warning that OpenAI’s new browser, ChatGPT Atlas, could be vulnerable to malicious attacks that could turn AI assistants against users, potentially stealing sensitive data or even draining their bank accounts. The AI company launched Atlas on Tuesday, with the goal of…
OpenAI has launched its own browser, Atlas, which has both ChatGPT and AI agents built in. But there are also serious risks in letting a chatbot do all sorts of things on your behalf, the warning says.
Coverage Details
Total News Sources18
Leaning Left1Leaning Right1Center3Last UpdatedBias Distribution60% Center
Bias Distribution
- 60% of the sources are Center
60% Center
L 20%
C 60%
R 20%
Factuality
To view factuality data please Upgrade to Premium
